Kyle L. Crandall

University of Utah, George Washington University

Papers

Total Citations

H-Index

About

Kyle L. Crandall’s research lies at the intersection of robotics, state estimation, and multi-agent systems, with a focus on enabling robots to operate more intelligently under uncertainty. His most influential work, “Online parameter estimation via real-time replanning of continuous Gaussian POMDPs” (2014, 19 citations), addresses a fundamental challenge in robotics: accurately modeling a robot’s dynamics. Crandall developed a method that simultaneously estimates unknown physical parameters—such as mass or moment of inertia—while planning actions in real time, using a Partially Observable Markov Decision Process (POMDP) framework. This approach allows robots to adapt their models on the fly, improving performance in tasks like motion planning and control without requiring exhaustive pre-calibration. In a different vein, his work “Using abstraction for swarm control of a parent system” (2016) explores how large robot swarms can cooperatively manipulate a single “parent” system with its own dynamics, introducing abstraction techniques to manage the complexity of decentralized control. Though early in his career, Crandall’s contributions demonstrate a clear commitment to making autonomous systems more robust and adaptive, bridging theoretical planning algorithms with practical robotic applications.